huggingface-vision-trainer

name	huggingface-vision-trainer
description	Codex curated Hugging Face skill for training or fine-tuning vision models with image datasets, transforms, metrics, checkpoints, and Hub integration.
tags	["codex-builtin","huggingface","vision","training","image-classification"]

huggingface-vision-trainer

When to Use

Use for image classification or vision model training pipelines, dataset transforms, evaluation metrics, and checkpoint publishing.

Do Not Use When

Do not use for text-only LLM fine-tuning or Gradio UI-only changes.

Required Inputs

Image dataset, model checkpoint, transforms, metrics, hardware constraints, and output repo or checkpoint path.

Workflow

1. Load and inspect image dataset splits.
2. Define transforms and trainer config.
3. Run a small smoke train/eval.
4. Save and optionally push the checkpoint.